Civil works, operation and maintenance of urban infrastructure

Abstract

Maintaining road infrastructure is of utmost importance for urban planning. Road construction projects are both dynamic and intricate, requiring the coordination of numerous workers, materials, equipment, and processes. Scan-to-BIM has significantly influenced urban planning, by offering precise documentation of as-built data and predictive maintenance. Recently, a variety of remote sensing technologies have been utilized to improve the productivity of the construction industry. While the demand of 3D geo-information is on the rise, LiDAR and navigation sensors parameters are undergoing rapid changes.To progress our existing road construction methods, it is necessary to enhance innovation, minimize construction-related uncertainties, and comprehensively plan the entire lifespan of the infrastructure. Scan-to-BIM provides a solid foundation for understanding urban environments. It enables the identification and analysis of various objects and features surrounding the roads, such as street furniture, buildings, vegetation, and landmarks. The main aim of this thesis was to develop automated data processing for monitoring and extracting relevant semantic information from the road and urban environment. These techniques are designed to monitor and extract meaningful semantic information from the road and urban environments, particularly for Scan-to-BIM applications. For the evaluation and analysis of Mobile Laser Scanning (MLS) systems, three distinct environments: roads, urban areas, and semi-urban areas are considered. Our study conducts a comprehensive comparison of three LiDAR systems, employing different machine learning classifiers. The experimental results demonstrate that for highway mapping tasks, MLS-single head system excels due to its ability to generate dense point clouds, crucial for capturing intricate details. However, in urban and semi-urban environments with numerous occlusions and complex scenarios, the MLS-dual head system emerges as the most suitable option, effectively handling occlusions to ensure comprehensive data coverage. The findings presented here offer valuable insights for researchers and professionals seeking to select a mobile laser scanner for mapping transport infrastructure. In terms of classifier performance, Random Forest consistently proves to be the best choice. Additionally, our thesis introduces a novel contribution within the Scan-to-BIM framework, utilizing depth information derived from RGB images captured by on-board cameras on UAVs for creating depth images. UAV-based data acquisition is advantageous due to its portability, lightweight nature, cost-effectiveness, and wide coverage compared to MLS. In conclusion, the analysis of MLS systems and UAVs and their performance evaluation across different environments aids decision-making processes in the field of intelligent transport infrastructure mapping, paving the way for automatic detection and identification of crucial information, including geometric characteristics and semantic properties of linear elements and traffic-related features. This dissertation, recognized with an Industrial mention in collaboration with Igeniería Insitu, offers practical solutions, informed decision-making tools, and state-of-the-art scanning technologies suitable for various road transport infrastructure projects. By bridging the gap between academia and industry, this dissertation not only contributes to the academic knowledge base but also directly benefits the community by enhancing the efficiency and quality of our road infrastructure, making it a key component of any Intelligent Infrastructure system. Therefore, the research conducted in this dissertation involves real case studies,contributes to the solution of the above-mentioned methodologies, and advances the current state-of the art.